Category Archives: Big Data

Artificial Intelligence for Monitoring Elections (AIME)

AIME logo

I published a blog post with the same title a good while back. Here’s what I wrote at the time:

Citizen-based, crowdsourced election observation initiatives are on the rise. Leading election monitoring organizations are also looking to leverage citizen-based reporting to complement their own professional election monitoring efforts. Meanwhile, the information revolution continues apace, with the number of new mobile phone subscriptions up by over 1 billion in just the past 36 months alone. The volume of election-related reports generated by “the crowd” is thus expected to grow significantly in the coming years. But international, national and local election monitoring organizations are completely unprepared to deal with the rise of Big (Election) Data.

I thus introduced a new project to “develop a free and open source platform to automatically filter relevant election reports from the crowd.” I’m pleased to report that my team and I at QCRI have just tested AIME during an actual election for the very first time—the 2015 Nigerian Elections. My QCRI Research Assistant Peter Mosur (co-author of this blog post) collaborated directly with Oludotun Babayemi from Clonehouse Nigeria and Chuks Ojidoh from the Community Life Project & Reclaim Naija to deploy and test the AIME platform.

AIME is a free and open source (experimental) solution that combines crowd-sourcing with Artificial Intelligence to automatically identify tweets of interest during major elections. As organizations engaged in election monitoring well know, there can be a lot chatter on social media as people rally behind their chosen candidates, announce this to the world, ask their friends and family who they will be voting for, and updating others when they have voted while posting about election related incidents they may have witnessed. This can make it rather challenging to find reports relevant to election monitoring groups.


Election monitors typically monitor instances of violence, election rigging, and voter issues. These incidents are monitored because they reveal problems that arise with the elections. Election monitoring initiatives such as Reclaim Naija & Uzabe also monitor several other type of incidents but for the purposes of testing the AIME platform, we selected three types of events mentioned above. In order to automatically identify tweets related to these events, one must first provide AIME with example tweets. (Of course, if there is no Twitter traffic to begin with, then there won’t be much need for AIME, which is precisely why we developed an SMS extension that can be used with AIME).

So where does the crowdsourcing comes in? Users of AIME can ask the crowd to tag tweets related to election-violence, rigging and voter issues by simply clicking on tagging tweets posted to the AIME platform with the appropriate event type. (Several quality control mechanisms are built in to ensure data quality. Also, one does not need to use crowdsourcing to tag the tweets; this can be done internally as well or instead). What AIME does next is use a technique from Artificial Intelligence (AI) called statistical machine learning to understand patterns in the human-tagged tweets. In other words, it begins to recognize which tweets belong in which category type—violence, rigging and voter issues. AIME will then auto-classify new tweets that are related to these categories (and can auto-classify around 2 millions tweets or text messages per minute).

Screen Shot 2015-04-10 at 8.33.08 AM

Before creating our automatic classifier for the Nigerian Elections, we first needed to collect examples of tweets related to election violence, rigging and voter issues in order to teach AIME. Oludotun Babayemi and Chuks Ojidoh kindly provided the expert local knowledge needed to identify the keywords we should be following on Twitter (using AIME). They graciously gave us many different keywords to use as well as a list of trusted Twitter accounts to follow for election-related messages. (Due to difficulties with AIME, we were not able to use the trusted accounts. In addition, many of the suggested keywords were unusable since words like “aggressive”, “detonate”, and “security” would have resulted in large amount of false positives).

Here is the full list of keywords used by AIME:

Nigeria elections, nigeriadecides, Nigeria decides, INEC, GEJ, Change Nigeria, Nigeria Transformation, President Jonathan, Goodluck Jonathan, Sai Buhari, saibuhari, All progressives congress, Osibanjo, Sambo, Peoples Democratic Party, boko haram, boko, area boys, nigeria2015, votenotfight, GEJwinsit, iwillvoteapc, gmb2015, revoda, thingsmustchange,  and march4buhari   

Out of this list, “NigeriaDecides” was by far the most popular keyword used in the elections. It accounted for over 28,000 Tweets of a batch of 100,000. During the week leading up to the elections, AIME collected roughly 800,000 Tweets. Over the course of the elections and the few days following, the total number of collected Tweets jumped to well over 4 million.

We sampled just a handful of these tweets and manually tagged those related to violence, rigging and other voting issues using AIME. “Violence” was described as “threats, riots, arming, attacks, rumors, lack of security, vandalism, etc.” while “Election Rigging” was described as “Ballot stuffing, issuing invalid ballot papers, voter impersonation, multiple voting, ballot boxes destroyed after counting, bribery, lack of transparency, tampered ballots etc.” Lastly, “Voting Issues” was defined as “Polling station logistics issues, technical issues, people unable to vote, media unable to enter, insufficient staff, lack of voter assistance, inadequate voting materials, underage voters, etc.”

Any tweet that did not fall into these three categories was tagged as “Other” or “Not Related”. Our Election Classifiers were trained with a total of 571 human-tagged tweets which enabled AIME to automatically classify well over 1 million tweets (1,263,654 to be precise). The results in the screenshot below show accurate AIME was at auto-classifying tweets based on the different event types define earlier. AUC is what captures the “overall accuracy” of AIME’s classifiers.


AIME was rather good at correctly tagging tweets related to “Voting Issues” (98% accuracy) but drastically poor at tagging related to “Election Rigging” (0%). This is not AIME’s fault : ) since it only had 8 examples to learn from. As for “Violence”, the accuracy score was 47%, which is actually surprising given that AIME only had 14 human-tagged examples to learn from. Lastly, AIME did fairly well at auto-classifying unrelated tweets (accuracy of 86%).

Conclusion: this was the first time we tested AIME during an actual election and we’ve learned a lot in the process. The results are not perfect but enough to press on and experiment further with the AIME platform. If you’d like to test AIME yourself (and if you fully recognize that the tool is experimental and still under development, hence not perfect), then feel free to get in touch with me here. We have 2 slots open for testing. In the meantime, big thanks to my RA Peter for spearheading both this deployment and the subsequent research.

Crowdsourcing Point Clouds for Disaster Response

Point Clouds, or 3D models derived from high resolution aerial imagery, are in fact nothing new. Several software platforms already exist to reconstruct a series of 2D aerial images into fully fledged 3D-fly-through models. Check out these very neat examples from my colleagues at Pix4D and SenseFly:

What does a castle, Jesus and a mountain have to do with humanitarian action? As noted in my previous blog post, there’s only so much disaster damage one can glean from nadir (that is, vertical) imagery and oblique imagery. Lets suppose that the nadir image below was taken by an orbiting satellite or flying UAV right after an earthquake, for example. How can you possibly assess disaster damage from this one picture alone? Even if you had nadir imagery for these houses before the earthquake, your ability to assess structural damage would be limited.

Screen Shot 2015-04-09 at 5.48.23 AM

This explains why we also captured oblique imagery for the World Bank’s UAV response to Cyclone Pam in Vanuatu (more here on that humanitarian mission). But even with oblique photographs, you’re stuck with one fixed perspective. Who knows what these houses below look like from the other side; your UAV may have simply captured this side only. And even if you had pictures for all possible angles, you’d literally have 100’s of pictures to leaf through and make sense of.

Screen Shot 2015-04-09 at 5.54.34 AM

What’s that famous quote by Henry Ford again? “If I had asked people what they wanted, they would have said faster horses.” We don’t need faster UAVs, we simply need to turn what we already have into Point Clouds, which I’m indeed hoping to do with the aerial imagery from Vanuatu, by the way. The Point Cloud below was made only from single 2D aerial images.

It isn’t perfect, but we don’t need perfection in disaster response, we need good enough. So when we as humanitarian UAV teams go into the next post-disaster deployment and ask what humanitarians they need, they may say “faster horses” because they’re not (yet) familiar with what’s really possible with the imagery processing solutions available today. That obviously doesn’t mean that we should ignore their information needs. It simply means we should seek to expand their imaginations vis-a-vis the art of the possible with UAVs and aerial imagery. Here is a 3D model of a village in Vanuatu constructed using 2D aerial imagery:

Now, the title of my blog post does lead with the word crowdsourcing. Why? For several reasons. First, it takes some decent computing power (and time) to create these Point Clouds. But if the underlying 2D imagery is made available to hundreds of Digital Humanitarians, we could use this distributed computing power to rapidly crowdsource the creation of 3D models. Second, each model can then be pushed to MicroMappers for crowdsourced analysis. Why? Because having a dozen eyes scrutinizing one Point Cloud is better than 2. Note that for quality control purposes, each Point Cloud would be shown to 5 different Digital Humanitarian volunteers; we already do this with MicroMappers for tweets, pictures, videos, satellite images and of course aerial images as well. Each digital volunteer would then trace areas in the Point Cloud where they spot damage. If the traces from the different volunteers match, then bingo, there’s likely damage at those x, y and z coordinate. Here’s the idea:

We could easily use iPads to turn the process into a Virtual Reality experience for digital volunteers. In other words, you’d be able to move around and above the actual Point Cloud by simply changing the position of your iPad accordingly. This technology already exists and has for several years now. Tracing features in the 3D models that appear to be damaged would be as simple as using your finger to outline the damage on your iPad.

What about the inevitable challenge of Big Data? What if thousands of Point Clouds are generated during a disaster? Sure, we could try to scale our crowd-sourcing efforts by recruiting more Digital Humanitarian volunteers, but wouldn’t that just be asking for a “faster horse”? Just like we’ve already done with MicroMappers for tweets and text messages, we would seek to combine crowdsourcing and Artificial Intelligence to automatically detect features of interest in 3D models. This sounds to me like an excellent research project for a research institute engaged in advanced computing R&D.

I would love to see the results of this applied research integrated directly within MicroMappers. This would allow us to integrate the results of social media analysis via MicroMappers (e.g, tweets, Instagram pictures, YouTube videos) directly with the results of satellite imagery analysis as well as 2D and 3D aerial imagery analysis generated via MicroMappers.

Anyone interested in working on this?

Artificial Intelligence Powered by Crowdsourcing: The Future of Big Data and Humanitarian Action

There’s no point spewing stunning statistics like this recent one from The Economist, which states that 80% of adults will have access to smartphones before 2020. The volume, velocity and variety of digital data will continue to skyrocket. To paraphrase Douglas Adams, “Big Data is big. You just won’t believe how vastly, hugely, mind-bogglingly big it is.”


And so, traditional humanitarian organizations have a choice when it comes to battling Big Data. They can either continue business as usual (and lose) or get with the program and adopt Big Data solutions like everyone else. The same goes for Digital Humanitarians. As noted in my new book of the same title, those Digital Humanitarians who cling to crowdsourcing alone as their pièce de résistance will inevitably become the ivy-laden battlefield monuments of 2020.


Big Data comprises a variety of data types such as text, imagery and video. Examples of text-based data includes mainstream news articles, tweets and WhatsApp messages. Imagery includes Instagram, professional photographs that accompany news articles, satellite imagery and increasingly aerial imagery as well (captured by UAVs). Television channels, Meerkat and YouTube broadcast videos. Finding relevant, credible and actionable pieces of text, imagery and video in the Big Data generated during major disasters is like looking for a needle in a meadow (haystacks are ridiculously small datasets by comparison).

Humanitarian organizations, like many others in different sectors, often find comfort in the notion that their problems are unique. Thankfully, this is rarely true. Not only is the Big Data challenge not unique to the humanitarian space, real solutions to the data deluge have already been developed by groups that humanitarian professionals at worst don’t know exist and at best rarely speak with. These groups are already using Artificial Intelligence (AI) and some form of human input to make sense of Big Data.

Data digital flow

How does it work? And why do you still need some human input if AI is already in play? The human input, which can be via crowdsourcing or a few individuals is needed to train the AI engine, which uses a technique from AI called machine learning to learn from the human(s). Take AIDR, for example. This experimental solution, which stands for Artificial Intelligence for Disaster Response, uses AI powered by crowdsourcing to automatically identify relevant tweets and text messages in an exploding meadow of digital data. The crowd tags tweets and messages they find relevant and the AI engine learns to recognize the relevance patterns in real-time, allowing AIDR to automatically identify future tweets and messages.

As far as we know, AIDR is the only Big Data solution out there that combines crowdsourcing with real-time machine learning for disaster response. Why do we use crowdsourcing to train the AI engine? Because speed is of the essence in disasters. You need a crowd of Digital Humanitarians to quickly tag as many tweets/messages as possible so that AIDR can learn as fast as possible. Incidentally, once you’ve created an algorithm that accurately detects tweets relaying urgent needs after a Typhoon in the Philippines, you can use that same algorithm again when the next Typhoon hits (no crowd needed).

What about pictures? After all, pictures are worth a thousand words. Is it possible to combine artificial intelligence with human input to automatically identify pictures that show infrastructure damage? Thanks to recent break-throughs in computer vision, this is indeed possible. Take Metamind, for example, a new startup I just met with in Silicon Valley. Metamind is barely 6 months old but the team has already demonstrated that one can indeed automatically identify a whole host of features in pictures by using artificial intelligence and some initial human input. The key is human input since this is what trains the algorithms. The more human-generated training data you have, the better your algorithms.

My team and I at QCRI are collaborating with Metamind to create algorithms that can automatically detect infrastructure damage in pictures. The Silicon Valley start-up is convinced that we’ll be able to create a highly accurate algorithms if we have enough training data. This is where MicroMappers comes in. We’re already using MicroMappers to create training data for tweets and text messages (which is what AIDR uses to create algorithms). In addition, we’re already using MicroMappers to tag and map pictures of disaster damage. The missing link—in order to turn this tagged data into algorithms—is Metamind. I’m excited about the prospects, so stay tuned for updates as we plan to start teaching Metamind’s AI engine this month.

Screen Shot 2015-03-16 at 11.45.31 AM

How about videos as a source of Big Data during disasters? I was just in Austin for SXSW 2015 and met up with the CEO of WireWax, a British company that uses—you guessed it—artificial intelligence and human input to automatically detect countless features in videos. Their platform has already been used to automatically find guns and Justin Bieber across millions of videos. Several other groups are also working on feature detection in videos. Colleagues at Carnegie Melon University (CMU), for example, are working on developing algorithms that can detect evidence of gross human rights violations in YouTube videos coming from Syria. They’re currently applying their algorithms on videos of disaster footage, which we recently shared with them, to determine whether infrastructure damage can be automatically detected.

What about satellite & aerial imagery? Well the team driving DigitalGlobe’s Tomnod platform have already been using AI powered by crowdsourcing to automatically identify features of interest in satellite (and now aerial) imagery. My team and I are working on similar solutions with MicroMappers, with the hope of creating real-time machine learning solutions for both satellite and aerial imagery. Unlike Tomnod, the MicroMappers platform is free and open source (and also filters social media, photographs, videos & mainstream news).

Screen Shot 2015-03-16 at 11.43.23 AM

Screen Shot 2015-03-16 at 11.41.21 AM

So there you have it. The future of humanitarian information systems will not be an App Store but an “Alg Store”, i.e, an Algorithm Store providing a growing menu of algorithms that have already been trained to automatically detect certain features in texts, imagery and videos that gets generated during disasters. These algorithms will also “talk to each other” and integrate other feeds (from real-time sensors, Internet of Things) thanks to data-fusion solutions that already exist and others that are in the works.

Now, the astute reader may have noted that I omitted audio/speech in my post. I’ll be writing about this in a future post since this one is already long enough.

This is How Social Media Can Inform UN Needs Assessments During Disasters

My team at QCRI just published their latest findings on our ongoing crisis computing and humanitarian technology research. They focused on UN/OCHA, the international aid agency responsible for coordinating humanitarian efforts across the UN system. “When disasters occur, OCHA must quickly make decisions based on the most complete picture of the situation they can obtain,” but “given that complete knowledge of any disaster event is not possible, they gather information from myriad available sources, including social media.” QCRI’s latest research, which also drew on multiple interviews, shows how “state-of-the-art social media processing methods can be used to produce information in a format that takes into account what large international humanitarian organizations require to meet their constantly evolving needs.”


QCRI’s new study (PDF) focuses specifically on the relief efforts in response to Typhoon Yolanda (known locally as Haiyan). “When Typhoon Yolanda struck the Philippines, the combination of widespread network access, high Twitter use, and English proficiency led to many located in the Philippines to tweet about the typhoon in English. In addition, outsiders located elsewhere tweeted about the situation, leading to millions of English-language tweets that were broadcast about the typhoon and its aftermath.”

When disasters like Yolanda occur, the UN uses the Multi Cluster/Sector Initial Rapid Assessment (MIRA) survey to assess the needs of affected populations. “The first step in the MIRA process is to produce a ‘Situation Analysis’ report,” which is produced within the first 48 hours of a disaster. Since the Situation Analysis needs to be carried out very quickly, “OCHA is open to using new sources—including social media communications—to augment the information that they and partner organizations so desperately need in the first days of the immediate post-impact period. As these organizations work to assess needs and distribute aid, social media data can potentially provide evidence in greater numbers than what individuals and small teams are able to collect on their own.”

My QCRI colleagues therefore analyzed the 2 million+ Yolanda-related tweets published between November 7-13, 2013 to assess whether any of these could have augmented OCHA’s situational awareness at the time. (OCHA interviewees stated that this “six-day period would be of most interest to them”). QCRI subsequently divided the tweets into two periods:

Screen Shot 2015-02-14 at 8.31.58 AM

Next, colleagues geo-located the tweets by administrative region and compared the frequency of tweets in each region with the number of people who were later found to have been affected in the respective region. The result of this analysis is displayed below (click to enlarge).

Screen Shot 2015-02-14 at 8.33.21 AM

While the “activity on Twitter was in general more significant in regions heavily affected by the typhoon, the correlation is not perfect.” This should not come as a surprise. This analysis is nevertheless a “worthwhile exercise, as it can prove useful in some circumstances.” In addition, knowing exactly what kinds of biases exist on Twitter, and which are “likely to continue is critical for OCHA to take into account as they work to incorporate social media data into future response efforts.”

QCRI researchers also analyzed the 2 million+ tweets to determine which  contained useful information. An informative tweet is defined as containing “information that helps you understand the situation.” They found that 42%-48% of the 2 million tweets fit this category, which is particularly high. Next, they classified those one million informative tweets using the Humanitarian Cluster System. The Up/Down arrows below indicate a 50%+ increase/decrease of tweets in that category during period 2.

Screen Shot 2015-02-14 at 8.35.53 AM

“In the first time period (roughly the first 48 hours), we observe concerns focused on early recovery and education and child welfare. In the second time period, these concerns extend to topics related to shelter, food, nutrition, and water, sanitation and hygiene (WASH). At the same time, there are proportionally fewer tweets regarding telecommunications, and safety and security issues.” The table above shows a “significant increase of useful messages for many clusters between period 1 and period 2. It is also clear that the number of potentially useful tweets in each cluster is likely on the order of a few thousand, which are swimming in the midst of millions of tweets. This point is illustrated by the majority of tweets falling into the ‘None of the above’ category, which is expected and has been shown in previous research.”

My colleagues also examined how “information relevant to each cluster can be further categorized into useful themes.” They used topic modeling to “quickly group thousands of tweets [and] understand the information they contain. In the future, this method can help OCHA staff gain a high- level picture of what type of information to expect from Twitter, and to decide which clusters or topics merit further examination and/or inclusion in the Situation Analysis.” The results of this topic modeling is displayed in the table below (click to enlarge).

Screen Shot 2015-02-14 at 8.34.37 AM

When UN/OCHA interviewees were presented with these results, their “feedback was positive and favorable.” One OCHA interviewee noted that this information “could potentially give us an indicator as to what people are talking most about— and, by proxy, apply that to the most urgent needs.” Another interviewee stated that “There are two places in the early hours that I would want this: 1) To add to our internal “one-pager” that will be released in 24-36 hours of an emergency, and 2) the Situation Analysis: [it] would be used as a proxy for need.” Another UN staffer remarked that “Generally yes this [information] is very useful, particularly for building situational awareness in the first 48 hours.” While some of the analysis may at times be too general, an OCHA interviewee “went on to say the table [above] gives a general picture of severity, which is an advantage during those first hours of response.”

As my QCRI team rightly notes, “This validation from UN staff supports our continued work on collecting, labeling, organizing, and presenting Twitter data to aid humanitarian agencies with a focus on their specific needs as they perform quick response procedures.” We are thus on the right track with both our AIDR and MicroMappers platforms. Our task moving forward is to use these platforms to produce the analysis discussed above, and to do so in near real-time. We also need to (radically) diversify our data sources and thus include information from text messages (SMS), mainstream media, Facebook, satellite imagery and aerial imagery (as noted here).

But as I’ve noted before, we also need enlightened policy making to make the most of these next generation humanitarian technologies. This OCHA proposal  on establishing specific social media standards for disaster response, and the official social media strategy implemented by the government of the Philippines during disasters serve as excellent examples in this respect.


Lots more on humanitarian technology, innovation, computing as well as policy making in my new book Digital Humanitarians: How Big Data is Changing the Face of Humanitarian Action.

Could This Be The Most Comprehensive Study of Crisis Tweets Yet?

I’ve been looking forward to blogging about my team’s latest research on crisis computing for months; the delay being due to the laborious process of academic publishing, but I digress. I’m now able to make their  findings public. The goal of their latest research was to “understand what affected populations, response agencies and other stakeholders can expect—and not expect—from [crisis tweets] in various types of disaster situations.”

Screen Shot 2015-02-15 at 12.08.54 PM

As my colleagues rightly note, “Anecdotal evidence suggests that different types of crises elicit different reactions from Twitter users, but we have yet to see whether this is in fact the case.” So they meticulously studied 26 crisis-related events between 2012-2013 that generated significant activity on twitter. The lead researcher on this project, my colleague & friend Alexandra Olteanu from EPFL, also appears in my new book.

Alexandra and team first classified crisis related tweets based on the following categories (each selected based on previous research & peer-reviewed studies):

Screen Shot 2015-02-15 at 11.01.48 AM

Written in long form: Caution & Advice; Affected Individuals; Infrastructure & Utilities; Donations & Volunteering; Sympathy & Emotional Support, and Other Useful Information. Below are the results of this analysis sorted by descending proportion of Caution & Advice related tweets (click to enlarge).

Screen Shot 2015-02-15 at 10.59.55 AM

The category with the largest number of tweets is “Other Useful Info.” On average 32% of tweets fall into this category (minimum 7%, maximum 59%). Interestingly, it appears that most crisis events that are spread over a relatively large geographical area (i.e., they are diffuse), tend to be associated with the lowest number of “Other” tweets. As my QCRI rightly colleagues note, “it is potentially useful to know that this type of tweet is not prevalent in the diffused events we studied.”

Tweets relating to Sympathy and Emotional Support are present in each of the 26 crises. On average, these account for 20% of all tweets. “The 4 crises in which the messages in this category were more prevalent (above 40%) were all instantaneous disasters.” This finding may imply that “people are more likely to offer sympathy when events […] take people by surprise.”

On average, 20% of tweets in the 26 crises relate to Affected Individuals. “The 5 crises with the largest proportion of this type of information (28%–57%) were human-induced, focalized, and instantaneous. These 5 events can also be viewed as particularly emotionally shocking.”

Tweets related to Donations & Volunteering accounted for 10% of tweets on average. “The number of tweets describing needs or offers of goods and services in each event varies greatly; some events have no mention of them, while for others, this is one of the largest information categories. “

Caution and Advice tweets constituted on average 10% of all tweets in a given crisis. The results show a “clear separation between human-induced hazards and natural: all human induced events have less caution and advice tweets (0%–3%) than all the events due to natural hazards (4%–31%).”

Finally, tweets related to Infrastructure and Utilities represented on average 7% of all tweets posted in a given crisis. The disasters with the highest number of such tweets tended to be flood situations.

In addition to the above analysis, Alexandra et al. also categorized tweets by their source:

Screen Shot 2015-02-15 at 11.23.19 AM

The results depicted below (click to enlarge) are sorted by descending order of eyewitness tweets.

Screen Shot 2015-02-15 at 11.27.57 AM

On average, about 9% of tweets generated during a given crises were written by Eyewitnesses; a figure that increased to 54% for the haze crisis in Singapore. “In general, we find a larger proportion of eyewitness accounts during diffused disasters caused by natural hazards.”

Traditional and/or Internet Media were responsible for 42% of tweets on average. ” The 6 crises with the highest fraction of tweets coming from a media source (54%–76%) are instantaneous, which make “breaking news” in the media.

On average, Outsiders posted 38% of the tweets in a given crisis while NGOs were responsible for about 4% of tweets and Governments 5%. My colleagues surmise that these low figures are due to the fact that both NGOs and governments seek to verify information before they release it. The highest levels of NGO and government tweets occur in response to natural disasters.

Finally, Businesses account for 2% of tweets on average. The Alberta floods of 2013 saw the highest proportion (9%) of tweets posted by businesses.

All the above findings are combined and displayed below (click to enlarge). The figure depicts the “average distribution of tweets across crises into combinations of information types (rows) and sources (columns). Rows and columns are sorted by total frequency, starting on the bottom-left corner. The cells in this figure add up to 100%.”

Screen Shot 2015-02-15 at 11.42.39 AM

The above analysis suggests that “when the geographical spread [of a crisis] is diffused, the proportion of Caution and Advice tweets is above the median, and when it is focalized, the proportion of Caution and Advice tweets is below the median. For sources, […] human-induced accidental events tend to have a number of eyewitness tweets below the median, in comparison with intentional and natural hazards.” Additional analysis carried out by my colleagues indicate that “human-induced crises are more similar to each other in terms of the types of information disseminated through Twitter than to natural hazards.” In addition, crisis events that develop instantaneously also look the same when studied through the lens of tweets.

In conclusion, the analysis above demonstrates that “in some cases the most common tweet in one crisis (e.g. eyewitness accounts in the Singapore haze crisis in 2013) was absent in another (e.g. eyewitness accounts in the Savar building collapse in 2013). Furthermore, even two events of the same type in the same country (e.g. Typhoon Yolanda in 2013 and Typhoon Pablo in 2012, both in the Philippines), may look quite different vis-à-vis the information on which people tend to focus.” This suggests the uniqueness of each event.

“Yet, when we look at the Twitter data at a meta-level, our analysis reveals commonalities among the types of information people tend to be concerned with, given the particular dimensions of the situations such as hazard category (e.g. natural, human-induced, geophysical, accidental), hazard type (e.g. earth-quake, explosion), whether it is instantaneous or progressive, and whether it is focalized or diffused. For instance, caution and advice tweets from government sources are more common in progressive disasters than in instantaneous ones. The similarities do not end there. When grouping crises automatically based on similarities in the distributions of different classes of tweets, we also realize that despite the variability, human-induced crises tend to be more similar to each other than to natural hazards.”

Needless to say, these are exactly the kind of findings that can improve the way we use MicroMappers & other humanitarian technologies for disaster response. So if want to learn more, the full study is available here (PDF). In addition, all the Twitter datasets used for the analysis are available at CrisisLex. If you have questions on the research, simply post them in the comments section below and I’ll ask my colleagues to reply there.


In the meantime, there is a lot more on humanitarian technology and computing in my new book Digital Humanitarians. As I note in said book, we also need enlightened policy making to tap the full potential of social media for disaster response. Technology alone can only take us so far. If we don’t actually create demand for relevant tweets in the first place, then why should social media users supply a high volume of relevant and actionable tweets to support relief efforts? This OCHA proposal on establishing specific social media standards for disaster response, and this official social media strategy developed and implemented by the Filipino government are examples of what enlightened leadership looks like.

Aerial Imagery Analysis: Combining Crowdsourcing and Artificial Intelligence

MicroMappers combines crowdsourcing and artificial intelligence to make sense of “Big Data” for Social Good. Why artificial intelligence (AI)? Because regular crowdsourcing alone is no match for Big Data. The MicroMappers platform can already be used to crowdsource the search for relevant tweets as well as pictures, videos, text messages, aerial imagery and soon satellite imagery. The next step is therefore to add artificial intelligence to this crowdsourced filtering platform. We have already done this with tweets and SMS. So we’re now turning our attention to aerial and satellite imagery.

Our very first deployment of MicroMappers for aerial imagery analysis was in Africa for this wildlife protection project. We crowdsourced the search for wild animals in partnership with rangers from the Kuzikus Wildlife Reserve based in Namibia. We were very pleased with the results, and so were the rangers. As one of them noted: “I am impressed with the results. There are at times when the crowd found animals that I had missed!” We were also pleased that our efforts caught the attention of CNN. As noted in that CNN report, our plan for this pilot was to use crowdsourcing to find the wildlife and to then combine the results with artificial intelligence to develop a set of algorithms that can automatically find wild animals in the future.

To do this, we partnered with a wonderful team of graduate students at EPFL, the well known polytechnique in Lausanne, Switzerland. While these students were pressed for time due to a number of deadlines, they were nevertheless able to deliver some interesting results. Their applied, computer vision research is particularly useful given our ultimate aim: to create an algorithm that can learn to detect features of interest in aerial and satellite imagery in near real-time (as we’re interested in applying this to disaster response and other time-sensitive events). For now, however, we need to walk before we can run. This means carrying out the tasks of crowdsourcing and artificial intelligence in two (not-yet-integrated) steps.

MM Oryx

As the EPFL students rightly note in their preliminary study, the use of thermal imaging (heat detection) to automatically identify wildlife in the bush is some-what problematic since “the temperature difference between animals and ground is much lower in savannah […].” This explains why the research team used the results of our crowdsourcing efforts instead. More specifically, they focused on automatically detecting the shadows of gazelles and ostriches by using an object based support vector machine (SVM). The whole process is summarized below.

Screen Shot 2015-02-09 at 12.46.38 AM

The above method produces results like the one below (click to enlarge). The circles represents the objects used to train the machine learning classifier. The discerning reader will note that the algorithm has correctly identified all the gazelles save for one instance in which two gazelles were standing close together were identified as one gazelle. But no other objects were mislabeled as a gazelle. In other words, EPFL’s gazelle algorithm is very accurate. “Hence the classifier could be used to reduce the number of objects to assess manually and make the search for gazelles faster.” Ostriches, on the other hand, proved more difficult to automatically detect. But the students are convinced that this could be improved if they had more time.

Screen Shot 2015-02-09 at 12.56.17 AM

In conclusion, more work certainly needs to be done, but I am pleased by these preliminary and encouraging results. In addition, the students at EPFL kindly shared some concrete features that we can implement on the MicroMappers side to improve the crowdsourced results for the purposes of developing automated algorithms in the future. So a big thank you to Briant, Millet and Rey for taking the time to carry out the above research. My team and I at QCRI very much look forward to continuing our collaboration with them and colleagues at EPFL.

In the meantime, more on all this in my new bookDigital Humanitarians: How Big Data is Changing the Face of Humanitarian Response, which has already been endorsed by faculty at Harvard, MIT, Stanford, Oxford, etc; and by experts at the UN, World Bank, Red Cross, Twitter, etc.

Video: Digital Humanitarians & Next Generation Humanitarian Technology

How do international humanitarian organizations make sense of the “Big Data” generated during major disasters? They turn to Digital Humanitarians who craft and leverage ingenious crowdsourcing solutions with trail-blazing insights from artificial intelligence to make sense of vast volumes of social media, satellite imagery and even UAV/aerial imagery. They also use these “Big Data” solutions to verify user-generated content and counter rumors during disasters. The talk below explains how Digital Humanitarians do this and how their next generation humanitarian technologies work.

Many thanks to TTI/Vanguard for having invited me to speak. Lots more on Digital Humanitarians in my new book of the same title.


Videos of my TEDx talks and the talks I’ve given at the White House, PopTech, Where 2.0, National Geographic, etc., are all available here.